Low soil moisture or a low water table removes peat’s water protection.
Indonesia Wildfire Evidence Report
How the evidence was assembled and constrained
Use this page to audit where the data came from, which preparation gates passed, how missingness was protected, and which conclusions remain outside the completed design.
Data sources represented in the research bundle
Source marks identify the provider for each evidence stream. They are attribution only, not endorsements, and availability does not make two sources directly comparable.
Source marks identify the datasets represented in this bundle; they are attribution only, not endorsements.
Global Forest Watch / WRIpeatland baseline
GWIS / JRCburned-area context
SiPongi / KemenhutIndonesia hotspot portalCHIRPS · ERA5-Land · AlphaEarthrainfall, fire weather, and prior-year satellite embeddingsWhich peatland conditions were tested?
The primary root-zone dryness interaction is complete and inconclusive. VPD, wind, and low EVI also remain inconclusive after the registered secondary-family correction; drainage still requires a separately timed analysis.
Canals and ditches can make peat dry, especially near disturbed edges.
Low pre-fire NDMI/EVI can indicate drier live fuel and degraded cover.
Rainfall deficit, high VPD, wind, and an ignition source can align in the same window.
Data preparation and validation audit
Daily coverage: 1,683 / 1,683 registered July–November days (100.0%).
All required environmental gates passed before the result above was fitted from the locked frame. Evidence for human access, intent, plantation development, and governance remains insufficient for attribution.
Data provenance and safeguards
Every browser-visible value comes from the frozen aggregate evidence bundle generated by the research pipeline.
See the source-specific limitation in the method.
Open source record ↗Aggregate burned-area/fire-count rows are descriptive only. They are not VIIRS overpass opportunity data, individual event onsets, or forest-only burned area.
Open source record ↗Positive portal records only; no swath coverage, valid non-detections, validated UTC, forest-only mask, or individual fire-event linkage. This is not a substitute for the VNP14IMG/VNP03IMG primary outcome.
Open source record ↗Positive portal records only; no swath coverage, valid non-detections, validated UTC, forest-only mask, or individual fire-event linkage. This is not a substitute for the VNP14IMG/VNP03IMG primary outcome.
Open source record ↗What the completed evidence does not establish
- The central human-accessibility and land-transformation hypothesis is not identifiable because the required matched-overpass exposure data are missing or unvalidated.
- GWIS values are aggregate historic four-province admin-1 burned-area rows, not individual event onsets, forest-only area, or a valid observation denominator.
- The GWIS legacy Kalimantan Timur unit includes territory now represented as Kalimantan Utara. It is not directly comparable with the current five-province SiPongi geography.
- SiPongi values are positive portal hotspot records, not fires, ignitions, unique events, occurrence rates, or detection probabilities.
- SiPongi all-platform counts cannot be interpreted longitudinally because portal platform composition changes; NASA-MODIS is the default display stratum.
- SiPongi 2024 is excluded: provider responses were quarantined after geography validation failed. Missing or rejected data are not rendered as zero.
- No raw SiPongi coordinates, local administrative labels, timestamps, or raw provider records are included in this explorer.
- The interactive globe uses rounded provincial aggregate reference anchors and simplified regional outlines for orientation only; it does not assert a vetted province-boundary geometry.
Evidence groups not sufficient for attribution: viirs_snpp_active_fire_and_geolocation, era5_land, chirps_v3, mapbiomas_indonesia_collection_4_1, mod13q1_061, historical_access_assets, peat_and_drainage_assets.